Central data lakes and warehouses concentrate all analytics data in one team that often lacks domain context, creating bottlenecks and poor data quality. Data mesh proposes a socio-technical shift: decentralise data ownership to the business domains that produce and understand it, and treat their data as a product for others to consume.
Four principles
Data mesh rests on four pillars working together. Domain ownership — each domain owns its analytical data. Data as a product — that data is discoverable, trustworthy and well-documented, with an owner accountable for quality. Self-serve data platform — a shared platform gives domains the tools to build and serve products without deep infra expertise. Federated computational governance — global standards (security, interoperability) are enforced automatically across autonomous domains.
| Aspect | Central lake/warehouse | Data mesh |
|---|---|---|
| Ownership | Central data team | Business domains |
| Scaling | Bottleneck | Scales with domains |
| Quality | Detached from context | Owned by domain experts |
| Governance | Central | Federated, automated |
Common pitfallData mesh is as much an organisational change as a technical one; without genuine domain ownership and product thinking it degrades into a rebranded data lake.
Applications
- Large enterprises with many data-producing domains
- Organisations whose central data team is a bottleneck
- Federated analytics across business units
References & further reading
- Dehghani, “Data Mesh: Delivering Data-Driven Value at Scale,” O'Reilly, 2022.
- Dehghani, “How to Move Beyond a Monolithic Data Lake to a Distributed Data Mesh,” martinfowler.com, 2019.
- Machado et al., “Data Mesh: Concepts and Principles,” 2022.